Direct Determination of the Interleukin-6 Binding Epitope of the Interleukin-6 Receptor by NMR Spectroscopy*

All cytokines belonging to the interleukin-6 (IL-6)-type family of cytokines utilize receptors that have a modular build of several immunoglobulin-like and fibronectin type III-like domains. Characteristic of these receptors is a cytokine receptor homology region consisting of two such fibronectin domains defined by a set of four conserved cysteines and a tryptophan-serine-X-tryptophan-serine sequence motif. On target cells, interleukin-6 first binds to its specific receptor and subsequently to a homodimer of the signal transducer protein gp130. The interleukin-6 receptor consists of three extracellular domains. The N-terminal immunoglobulin-like domain is not involved in ligand binding, whereas the third membrane proximal fibronectin-like domain accounts for more than 90% of the binding energy to IL-6. Here, the key residues of this fibronectin-like domain involved in the interaction with IL-6 are described. Chemical shift mapping data with 15N-labeled IL-6R-D3 and unlabeled IL-6 coupled with recent structural data clearly reveal the epitope within the IL-6R-D3 responsible for mediating the high affinity interaction with its cognate cytokine.

Cytokines are key mediators in the regulation and coordination of immune responses and hematopoiesis (1). Such molecules act on their target cells by binding to specific cell surface receptors and thereby induce receptor oligomerization (2). Cytokine receptors are type I transmembrane proteins. The receptors ectodomains consist of a number of fibronectin (FN) 1 type III-like and immunoglobulin-like (Ig) domains (3). The cytoplasmic region of these receptors lack intrinsic protein kinase activity but are constitutively associated with tyrosine kinases of the Janus kinase family (2,4). Ligand binding enforces homo-or heterodimerization of receptor molecules lead-ing to activation of the associated kinases and initiation of cytoplasmic signaling cascades (2,3).
Interleukin-6 (IL-6) is a pleiotropic cytokine involved in hematopoiesis, regulation of immune responses, and the acute phase reaction (2,5). This cytokine belongs to the family of interleukin-6-type cytokines that includes interleukin-11 (IL-11), cardiotrophin-1 (CT-1), ciliary neurotrophic factor (CNTF), leukemia inhibitory factor (LIF), oncostatin M (OSM), and cardiotrophin like factor. All of these cytokines share the characteristic four-helix bundle protein fold (6). This family of proteins signals via receptor complexes, which contain at least one molecule of gp130, the common signal transducing protein of the IL-6-type family of cytokines (3). IL-6 and IL-11 act via a homodimer of gp130, whereas CNTF, CT-1, LIF, and OSM induce a signaling event via the heterodimerization of gp130 and LIFR. Alternatively, OSM recruits gp130 and the OSM receptor. On target cells, IL-6, IL-11, and CNTF first bind to their specific ␣-receptor subunits, IL-6R, IL-11R, and CNTFR, respectively, which are not involved in the intracellular signal transduction cascade (3). These membrane-bound ␣-receptors can be functionally replaced by their soluble forms, which lack transmembrane and cytoplasmic parts (7). Interestingly, it has been shown recently that CNTF not only interacts with CNTFR but also with the IL-6R (8).
All receptors of the IL-6-type cytokines belong to the cytokine receptor class I family characterized by the presence of at least one cytokine receptor homology (CRH) region consisting of two FN domains (6). The conserved CRH plays a central role in cytokine recognition by the receptor. The N-terminal FN domain of the CRH contains four conserved cysteine residues, whereas the C-terminal FN domain is characterized by a tryptophan-serine-X-tryptophan-serine (WSXWS) sequence motif. The extracellular part of the IL-6R consists of an Ig domain (D1) and one CRH (D2 and D3). It is the C-terminal domain of IL-6R and IL-11R of the corresponding CRH that are primarily responsible for ligand binding, since these domains have been shown to account for more than 90% of the binding energy (9,10). However, the IL-6⅐IL-6R-D3 complex is not able to associate with the signal transducer gp130 and as such is not able to elicit a signaling cascade (9). The individual structures of IL-6 (11, 12) as well as IL-6R (13) have been solved by x-ray crystallography or NMR spectroscopy. In addition, the crystal structure (3.65 Å resolution) of the hexameric IL-6⅐IL-6R⅐gp130 complex has recently been published (14). A number of amino acid residues of IL-6 and IL-6R involved at the interaction interface have been defined by mutational studies (15) from molecular modeling (16) and with limitations from the recently published crystal structure of the hexameric IL-6⅐IL-6R⅐gp130 complex (14). Using chemical shift mapping in combination * This work was supported by Grant SFB 415 from the Deutsche Forschungsgemeinschaft (to J. G. and S. R.-J.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
with the current structural data the amino acid residues of IL-6R that are key in the interaction with its cognate cytokine are detailed. The results confirm the conclusions drawn from previous mutagenesis studies and clearly indicate that structural rearrangement of the IL-6R-D3 is required to induce the correct interaction between this domain and its cognate cytokine. In contrast, the NMR results presented do not entirely translate to the recent crystal structure of the complex and indicate that such structural rearrangements are not fully accounted for in the current complex model.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Unlabeled interleukin-6 was expressed and purified as described previously (17). The third domain of IL-6R was expressed using the BL21(DE3)pLysS Escherichia coli strain, refolded, and purified as described previously (9). 15 N-Labeled and 15 N/ 13 C double-labeled IL-6R-D3 was prepared by growing the bacteria in M9 minimal medium using 15 NH 4 Cl (1 g/liter) and 13 C 6labeled glucose (2 g/liter) as the sole nitrogen and carbon sources, respectively.
NMR Sample Preparation-After purification protein concentrations were determined by UV spectroscopy. Unlabeled IL-6 was mixed with the 15 15 N-enriched IL-6R-D3 was prepared in identical buffer conditions. All NMR data were acquired at 20°C on a Varian Inova Unity 600 MHz NMR spectrometer equipped with a z axis pulsed-field gradient 1 H/ 15 N/ 13 C probe optimized for 1 H detection. Two-dimensional 1 H-15 N HSQC experiments were recorded with the insensitive nuclei enhanced by polarization transfer delay set to 2.25 ms as a compromise between the fast relaxing amide protons and the optimal transfer time for the 1 J NH couplings of ϳ90 Hz. The recycle delay was 1.3 s. The data matrix consisted of 128* ϫ 384* data points (where n* refers to complex points) with acquisition times of 61 (t N ) and 96 ms (t HN ). A total of 32 scans per complex t N increment was collected. The total measuring time for each experiment was 1.6 h. The experiment was performed with the 1 H carrier positioned in the center of the amide proton frequency (7.52 ppm) and the 15 N carrier at 118.4 ppm. 15 N decoupling was applied during data acquisition. Sequence-specific backbone assignments of IL-6R-D3 were achieved 2 from a series of three-dimensional triple res-  13 C ␣ , and 13 C ␤ chemical shifts received an index value of ϩ1 or Ϫ1. For the 1 H␣ chemical shifts, three consecutive positive (downfield shift) or four consecutive negative (upfield shift) indices designate ␤-strand or ␣-helical structures, respectively. 13 C ␤ chemical shifts can only be used to identify stretches of ␤-strands, and a grouping of four or more positive indices denotes such a secondary structure (20,21). For the 1 H N and 13 C ␣ chemical shift values, three consecutive negative index values indicate that those residues form a ␤-strand, whereas four positive indices imply an ␣-helix (20,21). All other regions are designated as random coil or flexible regions of the protein. Residue numbers correspond to the full-length IL-6R.
The chemical shift data have been deposited in the BioMagResBank (www.bmrb.wisc.edu) data base, under accession number BMRB-5940.

RESULTS AND DISCUSSION
The two-dimensional 1 H-15 N HSQC spectrum of the IL-6R-D3 is presented in Fig. 1A. The sequence-specific 1 H-15 N assignment of IL-6R-D3 was derived from a series of threedimensional triple resonance experiments (see "Experimental Procedures") and resonances are labeled with assignment information. The chemical shift index (CSI) is a simple and reliable statistical method to predict secondary structure elements from recorded chemical shifts by comparison with their corresponding random coil values (20,21). From the calculated CSI values derived from the 1 H N , 1 H ␣ , 13 C ␣ , and 13 C ␤ chemical shifts a consensus CSI value can be determined that allows the identification of ␣-helical, ␤-strand, or coiled structures by indices of Ϫ1, ϩ1, or 0, respectively (20,21). The third domain of the extracellular part of IL-6R has been predicted to share a fibronectin type III-like fold consisting of seven ␤-strands that has been recently confirmed by the x-ray structure of the extracellular part of the IL-6R (13). The chemical shifts of 1 H N , 1 H ␣ , 15 N, 13 C ␣ , and 13 C ␤ nuclei of the IL-6R-D3 were derived from the following set of experiments: two-dimensional 1 H-15 N HSQC, three-dimensional HBHA(CBCACO)NH, three-dimensional CBCA(CO)NH, three-dimensional CBCANH, and threedimensional HNCA spectra. Chemical shift indexing was performed with the 1 H N , 1 H ␣ , 13 C ␣ , and 13 C ␤ chemical shifts to obtain an estimate of the secondary structure of the IL-6R-D3 (Fig. 1B). Of the expected seven ␤-strands observed in the crystal structure, six ␤-strands were estimated from the CSI values. Only the short ␤-strand G was not clearly identified. The CSI results agree with the lengths of the ␤-strands determined in the x-ray structure (13) with the exception of ␤-strand C that extends from Leu-232 to Ala-240 in the x-ray structure and from Leu-236 to Ala-240 in the IL-6R-D3. Apparently, the shorter ␤-strand C is due to the absence of assignment data for the 1 H ␣ and 13 C ␤ of Glu-235 due to spectral overlap.
The 1 H and 15 N chemical shifts of the resonances arising from the 15 N-enriched IL-6R-D3⅐unlabeled IL-6 complex were compared with the 1 H and 15 N chemical shifts of the resonances arising from the unbound 15 N-enriched IL-6R-D3. The majority of the resonances showed only minor or no chemical shift changes upon formation of the complex and were unambiguously assigned using the assignment data for the free IL-6R-D3. Such chemical shift changes are indicative that these residues of IL-6R-D3 are not located at the binding interface but are probably a reflection that binding of IL-6 causes subtle conformational changes to be transmitted through the protein to facilitate binding of IL-6. The much larger chemical shift changes or disappearance observed for some resonances abrogated the ability to unambiguously assign the full complement of resonances in the 1 H-15 N HSQC spectrum of IL-6R-D3 in complex with IL-6. The large shifts observed for these resonances are indicative of slow-to-intermediate exchange and strongly suggest that the binding of IL-6 to IL-6R-D3 has a low dissociation constant. This observation is supported by the measured k a and k d values of 2.72 ϫ 10 4 M Ϫ1 s Ϫ1 and 1.04 ϫ 10 Ϫ2 s Ϫ1 , respectively (9). Although we were not able to reassign all the resonances that showed the largest shifts or disappeared, due to intermediate exchange, using the assignment data for the free IL-6R-D3 did provide unambiguous identification of those key residues located at the binding interface (see below for discussion).
For a quantitative analysis of these data the normalized chemical shift differences were calculated. Fig. 2 shows the normalized chemical shift change for each residue of the IL-6R-D3 after binding to IL-6. Depending on the magnitude of the detected changes in the chemical shifts, residues were divided into two groups. The first group contains peaks that showed no or small changes in chemical shifts (Ͻ0.03 ppm). The second group contains peaks with a significantly larger chemical shift difference (Ͼ0.03 ppm). Chemical shift differences could not be calculated for some of the IL-6R-D3 residues, since their corresponding resonances in the 1 H-15 N HSQC spectra of the protein complex could not be unambiguously assigned. Although chemical shift differences were not calculated for these resonances, the residues were considered to be involved in the binding interface with IL-6. For some resonances (e.g. Asp-253 and Gln-272), nearby or overlapping peaks obscured accurate chemical shift measurements and were excluded from this analysis.
Using the crystal structure of the IL-6R (13), the measured changes in the 1 H and 15 N chemical shifts of the amide groups were projected onto the structure of the third extracellular domain of this receptor with regions colored red indicating the largest changes observed (i.e. Ͼ0.03 ppm). As can be seen in Fig. 3A, changes in IL-6R-D3 conformation upon interacting with IL-6 occur predominately in the B/C, D/E, and F/G loops and the adjacent stretches in the corresponding ␤-strands. Interestingly, some significant chemical shift changes were observed on the opposite side of this domain in the E/F loop. Since a contact with the ligand can be excluded, this may indicate that upon binding IL-6 a slight structural displacement of one or both of the ␤-strands E and F results in a subsequent movement of the E/F loop to accommodate the displacement. The involvement of the B/C, D/E, and F/G loops in the binding interface with IL-6 has been predicted from model building and site-directed mutagenesis studies (15,16,22). These predictions are confirmed by the NMR data.
Recently, the medium-resolution x-ray data of the IL-6⅐IL-6R⅐gp130 complex (14) coupled with high resolution structures of the single components, IL-6 (11), IL-6R (13), and the three membrane-distal domains of gp130 (23) were used to create a model of the protein complex. Using this model, the authors suggested several residues in IL-6R to be involved in the interaction with IL-6. In Fig. 3B the crystal structure of the IL-6⅐IL-6R (D2 and D3) complex is shown. The crucial role of the residues Phe-229 and Tyr-230 in the B/C loop and Glu-278 and Phe-279 in the F/G loop for ligand binding has been demonstrated by site-directed mutagenesis (15) and was confirmed by the structure of the IL-6⅐IL-6R complex (14). According to this model Phe-229 (B/C loop) and Phe-279 (F/G loop) contribute 28 and 20%, respectively, to the total binding area (14).
Besides Phe-229, an apparent conformational change of the IL-6R-D3 was also observed for the adjacent residues in the B/C loop upon binding to its cognate cytokine. Chemical shift differences greater than 0.03 ppm were observed between residues Trp-225 (0.040 ppm) and Arg-231 (0.037 ppm). The resonances arising from the amide groups of Asn-226 to Tyr-230 either broadened due to intermediate chemical exchange or disappeared and reappeared due to slow exchange in the spectrum and were consequently not assigned in the 1 H-15 N HSQC spectrum of the complex. Such changes clearly showed that these residues undergo conformational changes upon binding to IL-6. One of the significant differences between the crystal structure and our NMR data concerned Arg-233 (start of ␤-strand C). Arg-233 and the adjacent residues display chemical shift differences suggesting their involvement in IL-6 binding. In accordance with this observation are site-directed mutagenesis results that showed that Arg-233 plays an important role in binding to IL-6 (15). In the crystal structure, this residue apparently does not participate in the interaction with IL-6 but has intramolecular contacts with main-chain atoms of residue Phe-297, which is one of the signature residues in the IL-6/IL-6R interface (14). Therefore, the observed chemical shift difference of Arg-233 reflects a conformational change at the beginning of strand C rather than a direct involvement in ligand binding.
The chemical shift mapping data revealed residues located in the D/E loop (Fig. 3A) to be influenced by ligand binding. The amino acid residues in this loop, especially Leu-254, Gln-255, His-256, and in addition Cys-258 (beginning of ␤-strand E), FIG. 3. Display of residues of the third extracellular domain of the interleukin-6 receptor influenced strongly by binding of IL-6. A, the results of the chemical shift mapping data were translated onto the crystal structure of IL-6R-D3. ␤-Strands are labeled A-G, and residues that appear to undergo a significant structural change upon binding IL-6 are colored red. B, orientation of IL-6 and receptor IL-6R-D2D3 relative to each other as displayed in the crystal structure. Residues of IL-6R-D3 with chemical shift differences ⌬␦ Ͼ 0.03 ppm are marked in red. exhibit pronounced chemical shift differences (0.050, 0.048, 0.077, and 0.042, respectively), which are among the largest differences observed in this study. These observations are in excellent agreement with data derived from a site-directed mutagenesis studies (22). Yawata et al. (22) reported that a double mutant Gln-255-Leu and His-256-Gln resulted in a decreased binding capacity of 39% compared with the wild type protein. These data are somewhat contradictory to what can be deduced from the crystal structure of the complex (14). In this model, the D/E loop is far from the bound IL-6 with the shortest C␣-C␣ distance of 11.9 Å (IL-6, Lys-27; IL-6R-D3, Asp-253). Furthermore, the side chains of the corresponding residues of IL-6-D3 D/E loop are not geometrically positioned in the crystal structure to be making opposite contacts with the bound IL-6. Like in the case of the B/C loop, the pronounced chemical shift differences observed in the D/E loop seem to indicate a more general conformational change rather than a direct involvement in the interaction with IL-6.
Pronounced chemical shift differences were measured for resonances arising from residues located in the F/G loop (Fig. 3,  A and B). This loop and the adjacent ␤-strands comprise the residues Gln-276 to Glu-283. This region contains Phe-279, which, in analogy to the human growth hormone receptor, is the hot spot residue that contributes the largest ligand binding energy (24). The Phe-279 3 Ile mutation resulted in a complete loss of binding (22). For this residue as well as the preceding one (Glu-278), no resonances could be unambiguously identified in the 1 H-15 N HSQC spectra. The missing assignment data of Glu-278 and Phe-279 of IL-6R-D3 are indicative of fast chemical exchange (i.e. conformational flexibility) in the unbound form. Both residues have been shown to be crucial for IL-6 binding (15,22). This loop also contains two glycines, Gly-280 and Gly-282. Mutagenesis of Gly-282 abolished binding to IL-6 completely (22), reflecting the importance of this residue in the F/G loop. Since glycines are known to impart flexibility to loop regions, these two residues might also be responsible for the apparent flexibility of other residues in this loop. The measured chemical shift differences for the resonances corresponding to these two glycines are the largest measured (Fig. 2), indicating a distinct change in their chemical environment. These results imply an induced fit mechanism, in which the apparently flexible F/G loop adopts a specific rigid conformation upon IL-6 binding. Many cytokine receptors exhibit glycines at identical positions in the F/G loop (16) and the residues crucial for binding (hot spots) are also located in this loop (22, 24 -26). Intriguingly, in all type I cytokine receptors these conserved glycines precede the canonical WSXWS sequence motif that is the typical signature of the C-terminal domain of the CRH. It has been suggested that this motif plays a structural role (26 -28). Our data and other data suggest that this motif serves as an anchor for the adjacent mobile F/G loop that is crucial for the ligand receptor interaction.